Improving Performance of a Noise Reduction Algorithm by Switching the Analysis Filter Bank

Abstract

A new approach for preservation of transient parts of speech in a noise reduction system is proposed in this paper. Transient components of speech such as vowel onset and beginning of some consonants such as stop sounds are important parts for intelligibility of speech. These components are usually attenuated by noise reduction algorithms due to the low temporal resolution of block-based noise reduction techniques. A method is proposed to detect the transient component of speech, followed by dynamic switching of the analysis filter bank at the front end of the noise reduction system to provide higher resolution in the time domain. The optimal spectral gain values are transformed into the time domain to form a linear filter in order to achieve noise reduction and only group delay equalisation is performed to avoid discontinuity. Our objective evaluation shows that the proposed method provides superior performance compared to noise reduction with fixed time/frequency resolution analysis filter banks.

Adaptive Time-Frequency Analysis for Noise Reduction in an Audio Filter Bank With Low Delay

“…A number of adaptive time-frequency (TF) resolution schemes have been suggested, see for instance [8], [9] and the references therein. Also, it has been shown that adapting the TF resolution can lead to improvements in noise reduction [10], [11]. Common for these approaches is that they are not suitable for low delay implementations since they require longer time windows to determine the TF resolution and/or that they have a high computational complexity.…”